Storing 2 bytes in one
-
Hey, i've been looking for an effective method of storing 2 byte values within one. Or, two short values within one. The system I currently use is by embedding an unsafe code to point to a byte value and to cast the relevant data. Here's an example:
byte db, v1, v2;
v1 = 50;
v2 = 80;unsafe
{
byte* db_p = &db;
db_p[0] = (byte)((v1) >> 2);
db_p[1] = (byte)(((v1) << 6) | (((v2) >> 4)&0x3F));
db_p[2] = (byte)((v2) << 4);
}Which stores two values excellently. However, For values below 4 generally, numbers tend to overlap each other. IE, values of
v1 = 0, v2 = 0;andv1 = 1, v2 = 1;tend to have the same digit value ofdb. I was wondering there were any other methods of storing 2 values of data (bytes) within one single data (I'm using this to remove the necessity for multi-dimensional arrays and large collection objects). Thank you! -
Hey, i've been looking for an effective method of storing 2 byte values within one. Or, two short values within one. The system I currently use is by embedding an unsafe code to point to a byte value and to cast the relevant data. Here's an example:
byte db, v1, v2;
v1 = 50;
v2 = 80;unsafe
{
byte* db_p = &db;
db_p[0] = (byte)((v1) >> 2);
db_p[1] = (byte)(((v1) << 6) | (((v2) >> 4)&0x3F));
db_p[2] = (byte)((v2) << 4);
}Which stores two values excellently. However, For values below 4 generally, numbers tend to overlap each other. IE, values of
v1 = 0, v2 = 0;andv1 = 1, v2 = 1;tend to have the same digit value ofdb. I was wondering there were any other methods of storing 2 values of data (bytes) within one single data (I'm using this to remove the necessity for multi-dimensional arrays and large collection objects). Thank you!1. this isn't C# code is it? hence: wrong forum. 2. db_p is a pointer pointing to a byte, yet you write at locations db_p+0, db_p+1, db_p+2, so you write past the item your pointer points to, which is either not allowed (C#) or cheating (C/C++). 3. Of course you can store 2 byte values in 3 bytes; it is plain stupid. X|
Luc Pattyn [Forum Guidelines] [Why QA sucks] [My Articles]
I only read code that is properly indented, and rendered in a non-proportional font; hint: use PRE tags in forum messages
-
1. this isn't C# code is it? hence: wrong forum. 2. db_p is a pointer pointing to a byte, yet you write at locations db_p+0, db_p+1, db_p+2, so you write past the item your pointer points to, which is either not allowed (C#) or cheating (C/C++). 3. Of course you can store 2 byte values in 3 bytes; it is plain stupid. X|
Luc Pattyn [Forum Guidelines] [Why QA sucks] [My Articles]
I only read code that is properly indented, and rendered in a non-proportional font; hint: use PRE tags in forum messages
Well i'm sorry but i'm not entirely experienced with the whole "storing two values in one". I came here asking for help, not asking to be told i'm stupid and the methods i'm using are stupid.. That post didn't help at all, does anyone else have any real experience in storing data values inside one variable? Thanks
-
Hey, i've been looking for an effective method of storing 2 byte values within one. Or, two short values within one. The system I currently use is by embedding an unsafe code to point to a byte value and to cast the relevant data. Here's an example:
byte db, v1, v2;
v1 = 50;
v2 = 80;unsafe
{
byte* db_p = &db;
db_p[0] = (byte)((v1) >> 2);
db_p[1] = (byte)(((v1) << 6) | (((v2) >> 4)&0x3F));
db_p[2] = (byte)((v2) << 4);
}Which stores two values excellently. However, For values below 4 generally, numbers tend to overlap each other. IE, values of
v1 = 0, v2 = 0;andv1 = 1, v2 = 1;tend to have the same digit value ofdb. I was wondering there were any other methods of storing 2 values of data (bytes) within one single data (I'm using this to remove the necessity for multi-dimensional arrays and large collection objects). Thank you!No offense, but your code is odd. Now of course you can store multiple values in one variable, but they'll be smaller. Example:
static byte CombineBytes(byte a, byte b)
{
return (byte)((a & 15) | (b << 4));
}A and B can only be zero through 15 though, obviously. To get them back:
byte a = (byte)(x & 15);
byte b = (byte)(x >> 4);warning: all code here is untested. (but really it is just about the idea, not the actual code)
-
Hey, i've been looking for an effective method of storing 2 byte values within one. Or, two short values within one. The system I currently use is by embedding an unsafe code to point to a byte value and to cast the relevant data. Here's an example:
byte db, v1, v2;
v1 = 50;
v2 = 80;unsafe
{
byte* db_p = &db;
db_p[0] = (byte)((v1) >> 2);
db_p[1] = (byte)(((v1) << 6) | (((v2) >> 4)&0x3F));
db_p[2] = (byte)((v2) << 4);
}Which stores two values excellently. However, For values below 4 generally, numbers tend to overlap each other. IE, values of
v1 = 0, v2 = 0;andv1 = 1, v2 = 1;tend to have the same digit value ofdb. I was wondering there were any other methods of storing 2 values of data (bytes) within one single data (I'm using this to remove the necessity for multi-dimensional arrays and large collection objects). Thank you! -
Well, I plan on having a 2D plain with cells ranging from 15x15 to up to 300x300, and rather than using a library collection storing both values, I would prefer to have both values point to an index within an array. IE, R1C1 => Index 1, R1C2 => Index 2 etc etc.. I have read that these are possible but with no solutions.
-
No offense, but your code is odd. Now of course you can store multiple values in one variable, but they'll be smaller. Example:
static byte CombineBytes(byte a, byte b)
{
return (byte)((a & 15) | (b << 4));
}A and B can only be zero through 15 though, obviously. To get them back:
byte a = (byte)(x & 15);
byte b = (byte)(x >> 4);warning: all code here is untested. (but really it is just about the idea, not the actual code)
Thank you for this. Let's say I want to store a value greater than 15, would it be plausable to use a "short" value instead, to increase the available range of data? And, would each of the values produced be unique, or would they intersect at some point with conflicting values? Thank you for your help.
-
Well, I plan on having a 2D plain with cells ranging from 15x15 to up to 300x300, and rather than using a library collection storing both values, I would prefer to have both values point to an index within an array. IE, R1C1 => Index 1, R1C2 => Index 2 etc etc.. I have read that these are possible but with no solutions.
-
Thank you for this. Let's say I want to store a value greater than 15, would it be plausable to use a "short" value instead, to increase the available range of data? And, would each of the values produced be unique, or would they intersect at some point with conflicting values? Thank you for your help.
It just divided the byte in two parts, you could do the same with a short (then you could fit in two complete bytes) or an int (then you could fit in 2 complete shorts or 4 bytes) But if you want a 2D -> 1D mapping it's better to use something "sane" such as
x + y * widthor if you don't want to usewidthyou could use morton numbers to get a Z curve (just interleave the bits of your indices) -
Seriously? Do you mean:
index = x + y * widthThat's pretty much basic knowledge.. so I suspect you mean something else, but then what do you mean?Well, let's say each cell has a corresponding slot in an array containing information about the cell. Each cell, when clicked or managed, will refer to the single slot in an array corresponding to the co-ordinates or row,column values. If I was to use a method that relies on adding or multiplying x and y, there are chances that two cell coordinates may intersect and point to the same object. Therefore, I need the mathematical formula that produces an output using x and y to be a unique index key (not too large, otherwise the array would consume considerable amounts of memory), so that it may point to a slot in the array of cell information. 1,1 => Slot 1 in the array 1,2 => Slot 2 in the array 2,1 => Different slot than '1,2'. Thank you :)
-
Well, let's say each cell has a corresponding slot in an array containing information about the cell. Each cell, when clicked or managed, will refer to the single slot in an array corresponding to the co-ordinates or row,column values. If I was to use a method that relies on adding or multiplying x and y, there are chances that two cell coordinates may intersect and point to the same object. Therefore, I need the mathematical formula that produces an output using x and y to be a unique index key (not too large, otherwise the array would consume considerable amounts of memory), so that it may point to a slot in the array of cell information. 1,1 => Slot 1 in the array 1,2 => Slot 2 in the array 2,1 => Different slot than '1,2'. Thank you :)
-
I'm confused about what the value of 'width' should be. If width is the size of the cell, then I get conflicting types. When I ran a debug loop doing (x + y * 16) I received conflicts at 17,1 17,2 etc etc. And if I use 300 as the width, then the size of the array will be equal to 90,300 which is rediculously massive. Considering each slot in the array will store a structure of information, that could be quite a massive amount of memory usage?
-
I'm confused about what the value of 'width' should be. If width is the size of the cell, then I get conflicting types. When I ran a debug loop doing (x + y * 16) I received conflicts at 17,1 17,2 etc etc. And if I use 300 as the width, then the size of the array will be equal to 90,300 which is rediculously massive. Considering each slot in the array will store a structure of information, that could be quite a massive amount of memory usage?
The width of the 2D grid, if your coordinate goes up to 17 then either your width is >= 18 or your x coordinate it out of bounds (in that case it's only normal that it conflicts, it just wrapped to the next row) Edit: if you don't have a width (if your grid is essentially infinite) then you can still use morton numbers.
-
The width of the 2D grid, if your coordinate goes up to 17 then either your width is >= 18 or your x coordinate it out of bounds (in that case it's only normal that it conflicts, it just wrapped to the next row) Edit: if you don't have a width (if your grid is essentially infinite) then you can still use morton numbers.
I see, i'm sorry if I sound stupid i'm just not used to trying to store parameters like these. So, say I have a 300x300 map, would the formula be [x + y * 300]? Using this method, the value of the cell at 300 + 300 * 300 = 90,300. Would it be better to use bit-wise operations to try and reduce the total value down to a small index, or is this getting too complicated? If need be, I will simply have to make a collection of structures, and make each object key equal to "x,y". Although I would prefer to avoid using any forms of collections, and focus more on indexing values on the 2D map
-
I see, i'm sorry if I sound stupid i'm just not used to trying to store parameters like these. So, say I have a 300x300 map, would the formula be [x + y * 300]? Using this method, the value of the cell at 300 + 300 * 300 = 90,300. Would it be better to use bit-wise operations to try and reduce the total value down to a small index, or is this getting too complicated? If need be, I will simply have to make a collection of structures, and make each object key equal to "x,y". Although I would prefer to avoid using any forms of collections, and focus more on indexing values on the 2D map
Epoque wrote:
So, say I have a 300x300 map, would the formula be [x + y * 300]?
Yes. edit: note that the coordinate 300,300 is out of bounds in both directions, a 300x300 map only goes up to 299,299
Epoque wrote:
Would it be better to use bit-wise operations to try and reduce the total value down to a small index, or is this getting too complicated?
There is no reliable way to do that unless your grid is sparse in a specific and known way (in other words: probably not) If all your cells are used, the x + y * width formula will have an efficiency of 100% (it will number all w*h cells from 0 to w*h-1 with no gaps, so it can not be improved upon)
-
Well i'm sorry but i'm not entirely experienced with the whole "storing two values in one". I came here asking for help, not asking to be told i'm stupid and the methods i'm using are stupid.. That post didn't help at all, does anyone else have any real experience in storing data values inside one variable? Thanks
when I said "storing 2 byte values in 3 bytes is plain stupid." I can assure you it is, as that isn't compression at all, if anything it is expansion.
Epoque wrote:
i'm not entirely experienced with the whole "storing two values in one"
Nobody is, as it can't be done. If you could store two byte values in one byte (i.e. store two byte values and then retrieve those values in any order and as often as you'd like), then what would keep you from doing that recursively: you could store the first two bytes of anything in 1, the next two bytes in another one, then combine those two bytes into a final byte; in the end you could store any amount of data, why not the entire Internet, into that one final byte? We do get lunatics asking such questions regularly. OTOH one of the replies you posted earlier in this thread seems to indicate an entirely different problem: storing one (or N) values for a 2D collection of cells. That can easily be solved in many ways: 1. with N two-dimensional arrays 2. with N one-dimensional arrays, and linearizing the index (as in
index = x + y * width) 3. with a single one-dimensional array holding a little struct, which holds the different values for a single cell. If that is what you need, then you should have made that clear from the start. :)Luc Pattyn [Forum Guidelines] [Why QA sucks] [My Articles]
I only read code that is properly indented, and rendered in a non-proportional font; hint: use PRE tags in forum messages
-
Hey, i've been looking for an effective method of storing 2 byte values within one. Or, two short values within one. The system I currently use is by embedding an unsafe code to point to a byte value and to cast the relevant data. Here's an example:
byte db, v1, v2;
v1 = 50;
v2 = 80;unsafe
{
byte* db_p = &db;
db_p[0] = (byte)((v1) >> 2);
db_p[1] = (byte)(((v1) << 6) | (((v2) >> 4)&0x3F));
db_p[2] = (byte)((v2) << 4);
}Which stores two values excellently. However, For values below 4 generally, numbers tend to overlap each other. IE, values of
v1 = 0, v2 = 0;andv1 = 1, v2 = 1;tend to have the same digit value ofdb. I was wondering there were any other methods of storing 2 values of data (bytes) within one single data (I'm using this to remove the necessity for multi-dimensional arrays and large collection objects). Thank you!I don't see why you're using an unsafe context to do that. And you basically can't store sixteen bits in an eight-bit byte.
-
Well, I plan on having a 2D plain with cells ranging from 15x15 to up to 300x300, and rather than using a library collection storing both values, I would prefer to have both values point to an index within an array. IE, R1C1 => Index 1, R1C2 => Index 2 etc etc.. I have read that these are possible but with no solutions.
It seems you are essentially looking for a Hashing algorithm to take 2 values and create a single unique value where the hash would be unique across your entire set; for example the hash of the values 14, 16 would need to be different than 16, 14. I would recommend using one of the built-in collection classes that implement such algorithms rather than trying to create your own. The Hashtable and Dictionary classes should work nicely or you can use the System.Drawing.Point class to create coordinates. Personally, I'd use Dictionary where the Points are X, Y coordinates of the cells in the set. Since Point is a struct its operated on as a value type and can easily lend itself to what you're trying to do. I believe the default Point struct's X and Y members are int and if you are worried about wasted memory you could create your own Point struct that has X and Y of byte, or short, etc. If you decide to use the Dictionary class try to pass in the number of cells to the constructor (you will hopefully know that upon creation) and it will help performance slightly by preventing it from automatically growing its internal collection as items are added.
Hope in one hand and poop in the other; see which fills up first. Hope and change were good slogans, now show us more than words.
-
It seems you are essentially looking for a Hashing algorithm to take 2 values and create a single unique value where the hash would be unique across your entire set; for example the hash of the values 14, 16 would need to be different than 16, 14. I would recommend using one of the built-in collection classes that implement such algorithms rather than trying to create your own. The Hashtable and Dictionary classes should work nicely or you can use the System.Drawing.Point class to create coordinates. Personally, I'd use Dictionary where the Points are X, Y coordinates of the cells in the set. Since Point is a struct its operated on as a value type and can easily lend itself to what you're trying to do. I believe the default Point struct's X and Y members are int and if you are worried about wasted memory you could create your own Point struct that has X and Y of byte, or short, etc. If you decide to use the Dictionary class try to pass in the number of cells to the constructor (you will hopefully know that upon creation) and it will help performance slightly by preventing it from automatically growing its internal collection as items are added.
Hope in one hand and poop in the other; see which fills up first. Hope and change were good slogans, now show us more than words.
-
I don't see why you're using an unsafe context to do that. And you basically can't store sixteen bits in an eight-bit byte.
yet another non-believer. :doh:
Luc Pattyn [Forum Guidelines] [Why QA sucks] [My Articles]
I only read code that is properly indented, and rendered in a non-proportional font; hint: use PRE tags in forum messages